2
$\begingroup$

For $m, d\ge 1$, let $f_n: K\to \mathbb{R}^m$ be a continuous function for every $n\ge 0$, where $K\subset \mathbb{R}^d$ is compact. Show that if $\sum_{n=0}^\infty \|f_n\|_K <\infty$ then $f(x):=\sum_{n=0}^\infty f_n(x) <\infty$ is also a continuous function $f:K\to\mathbb{R}^m$.

Proof:

$\infty>\sum_{n=0}^\infty \|f_n\|_K =\sum_{n=0}^\infty \max\{\|f_n(x)\|:x\in K\} \ge \>\sum_{n=0}^\infty \|f_n(x)\|$, thus $\sum_{n=0}^\infty f_n(x)$ converges for any $x\in K$, which implies that $f(x)$ exists. This also implies that $\forall \varepsilon > 0, \exists N>0$, such that $k>N\implies \|\sum_{n=0}^k f_n(x) - f(x)\|<\varepsilon/2$.

Since $f_n(x)$ is continuous for every $x\in K$, then $\forall \varepsilon >0, \exists \delta > 0$ such that, for $x_0\in K$, $\|x-x_0\|<\delta\implies \|f_n(x)-f_n(x_0)\| < \varepsilon$.

Now, let $k>N$, then $\left\| \sum_{n=0}^k f_n(x)-\sum_{n=0}^k f_n(x_0)\right\| =\left\| \sum_{n=0}^k f_n(x)+f(x_0)-f(x_0)-\sum_{n=0}^k f_n(x_0)\right\|\le$ $\left\| \sum_{n=0}^k f_n(x)-f(x_0)\right\|+\left\|\sum_{n=0}^k f_n(x_0)-f(x_0)\right\|$, and this is where I get stuck since I do not know what to do with the $\left\| \sum_{n=0}^k f_n(x)-f(x_0)\right\|$ term, so some help would be appreciated. Is there something that I do not know about compact sets which can be applied here?

Update: I've just found out that sequences of continuous functions are uniformly convergent on compact sets, so I would change my last inequality as follows:

$\left\| \sum_{n=0}^k f_n(x)-\sum_{n=0}^k f_n(x_0)\right\| =\left\| \sum_{n=0}^k f_n(x)+f(x)-f(x)-\sum_{n=0}^k f_n(x_0)\right\|\le$ $\left\| \sum_{n=0}^k f_n(x)-f(x)\right\|+\left\|\sum_{n=0}^k f_n(x_0)-f(x)\right\| \le \varepsilon/2+\sup\limits_{x\in K}\left\|\sum_{n=0}^k f_n(x)-f(x)\right\|<\varepsilon/2+\varepsilon/2=\varepsilon$

Do you think this is correct now?

  • 0
    You mean *normally* convergent, not absolutely convergent.2017-02-16
  • 0
    Yes, "absolutely convergent" has a very unclear referent. Further, if that's what is meant, maybe "uniformly convergent" (or "uniformly on compacts") is better terminology than "normally"? The latter is not very descriptive, even if it does have a certain tradition behind it.2017-02-16

1 Answers 1

2

Since $\sum_{n=0}^\infty \|f_n\|_K < \infty$ for any $\epsilon > 0$ there exists $M \in \mathbb{N}$ such that

$$\sum_{n=M+1}^\infty \|f_n\|_K < \frac{\epsilon}{3},$$

and for every $x \in K$ we have

$$\begin{align}\left\|f(x) - \sum_{n=0}^M f_n(x)\right\| &= \left\|\sum_{n=M+1}^\infty f_n(x)\right\| \\ &\leqslant \sum_{n=M+1}^\infty \|f_n(x)\| \\ &\leqslant \sum_{n=M+1}^\infty \|f_n\|_K \\ &< \frac{\epsilon}{3}\end{align}.$$

Hence, if $x,x_0 \in K$ we have

$$\left\|f(x) - f(x_0) \right\| \leqslant \left\|f(x) - \sum_{n=0}^M f_n(x)\right\| + \left\|f(x_0) - \sum_{n=0}^M f_n(x_0)\right\| + \left\|\sum_{n=0}^M f_n(x) - \sum_{n=0}^M f_n(x_0)\right\| \\ < \frac{2\epsilon}{3} + \left\|\sum_{n=0}^Mf_n(x) - \sum_{n=0}^M f_n(x_0)\right\|. $$

Since $f_n$ is continuous on $K$ for every $n$, the finite sum $\sum_{n=0}^M f_n$ (where $M$ is fixed) is continuous at $x_0 \in K$. There exists $\delta > 0$ (which may depend on $x_0$ and $M$) such that if $\|x - x_0\| < \delta$ then

$$\left\|\sum_{n=0}^Mf_n(x) - \sum_{n=0}^M f_n(x_0)\right\| < \frac{\epsilon}{3},$$

and, consequently,

$$\left\|f(x) - f(x_0) \right\| < \epsilon.$$

Therefore $f$ is continuous at each point $x_0 \in K$. Note that compactness of $K$ was not needed to prove continuity. If, however, $K$ is compact then we have automatically uniform continuity of $f$ on $K.$

Regarding your update -- a sequence of continuous functions on a compact set need not be uniformly convergent. A counterexample is $f_n(x) = x^n$ where each $f_n$ is continuous on the compact interval $[0,1]$ but the sequence is not uniformly convergent.

  • 0
    Thanks! How can you see by inspection that $x^n$ is not uniformly convergent?2017-02-17
  • 2
    @sequence: For $f_n(x) = x^n$ we have $f_n(x) \to f(x)$ where $f(x) = 0$ for $0 \leqslant x < 1$ and $f(1) = 1$. The sequence of continuous functions cannot converge uniformly since the limit function is discontinuous. Any uniformly convergent sequence of continuous functions must converge to a continuous function. We just proved this in particular for a series (i.e., sequence of partial sums).2017-02-17
  • 0
    Actually I think my proof could be correct because we have absolute convergence of $\sum f_k$ on a compact set, so it is also uniformly convergent. @RRL2017-02-17
  • 0
    @sequence: This does not help in your approach. You still have the problem of how to estimate your term $\left\|\sum_{n=0}^k f_n(x_0)-f(x)\right\|$ where the partial sum and $f$ are evaluated at different points. You need to split $\|f(x) - f(x_0)\|$ into three terms that can be estimated as I have shown.2017-02-17
  • 0
    Why do you think the below would be incorrect? $\left\| \sum_{n=0}^k f_n(x)-\sum_{n=0}^k f_n(x_0)\right\| =\left\| \sum_{n=0}^k f_n(x)+f(x)-f(x)-\sum_{n=0}^k f_n(x_0)\right\|\le$ $\left\| \sum_{n=0}^k f_n(x)-f(x)\right\|+\left\|\sum_{n=0}^k f_n(x_0)-f(x)\right\| \le \varepsilon/2+\sup\limits_{x\in K}\left\|\sum_{n=0}^k f_n(x)-f(x)\right\|<\varepsilon/2+\varepsilon/2=\varepsilon$2017-02-17
  • 0
    How do you know as you claim that $\left\|\sum_{n=0}^k f_n(x_0)-f(x)\right\| < \epsilon/2$ for sufficiently large $k$? The partial sum here converges to $f(x_0)$ not $f(x)$. Even if you bound with $\left\|\sum_{n=0}^k f_n(x_0)-f(x_0)\right\| + \|f(x_0) - f(x) \|$ you need to argue that $\|f(x_0) - f(x)\| $can be made small -- which is using what you are trying to prove!2017-02-17
  • 0
    But $f(x_0)\le \sup\{f\}$, that's what I'm using in the inequality.2017-02-17